This study investigates the effect of hyperparameter optimization with grid search on the XGBoost and LightGBM models in
stroke classification. The research results show that parameter optimization significantly improves the performance of both
models, especially in accuracy, precision, F1 Score, and ROC-AUC Score. In the XGBoost model, improvements are mainly
seen in accuracy and precision, while LightGBM shows even improvements in all evaluation metrics. These findings underscore
the importance of hyperparameter optimization in building effective classification models to predict stroke risk more accurately
and reliably. These findings may contribute to further understanding of the factors that influence stroke and support more
appropriate and effective treatment in clinical practice